Problem Description: The focus and challenge of this project is to improve the quality of detection, biomass estimation and classification of fish and mammal species using machine learning (ML). For estimating biomass, acoustic sonar systems are preferable as their range far exceeds light-based system like cameras and lasers. For identification of (near range) objects, camera images may be better. In this work a sensor system (with a subsea camera pointed in the same direction as the sonar) able to combine both modalities is explored.
Proper classification of biomass and fish species is an important tool in sustainable harvesting of the ocean resources. The ability to correctly identify individual species and their size gives a possibility to monitor the ocean health status, and to identify events that leads to the disappearance of a species due to pollution, overfishing or other environmental issues.
During the work, algorithms for detecting size, class and proximity of sea animals in the camera images should be selected and implemented. The results from the camera detections should be used as a training and verification set when developing an algorithm for detecting comparable results in the acoustic data.
A large set of data for this thesis is already gathered during the environmental research project “Frisk Oslofjord”. The students will be also given the possibility to perform their own field operations and implement modifications in the data gathering.
Sketch for the project thesis
Contacts outside IDI
This is a project in collaboration with an external partner. If you choose this project, then I will serve as the responsible from NTNUs side, but the actual work will also be in tight collaboration with personell from the external partner as listed above.
If you consider picking a project with me as the supervisor, then please look at www.idi.ntnu.no/~helgel/teaching/proposals/.